Aengus Martin

A toolkit for designing interactive musical agents

We have developed a prototype software toolkit to enable non-technical users to design artificially intelligent agents to perform electronic music in collaboration with a human musician. In this paper we describe the toolkit and present a preliminary investigation of its use. We then discuss how the investigation has helped identify issues to address in an upcoming user-centred design study, which will take place in Spring 2011.

The agent designer toolkit

The Agent Designer Toolkit is the result of a study of methods for designing the behaviour of musical agents (i.e. autonomous systems) intended to perform high-level musical decision-making. It uses machine learning methods informed by a musicians knowledge and insights, to discover the salient musical patterns demonstrated in a set of example performances. Based on these patterns, it can produce agents with a variety of behaviours, corresponding to differing degrees of similarity to the demonstrated performance style. The agents can perform in real-time and respond to other musicians or external factors.

Design and Evaluation of Agents that Sequence and Juxtapose Short Musical Patterns in Real Time

We present and discuss the Agent Designer, a system that enables users of digital audio workstations to generate novel high-level structures for their compositions based on previous examples. The system uses variable-order Markov models and rule induction to learn both temporal relations and structural relations between parts in a piece of music. As is usual in machine learning, however, the quality of the learning can be improved greatly by users specifying relevant features. The Agent Designer therefore points to important design and human–computer interaction problems, as well as algorithmic challenges. We present a number of studies that help to understand how effective the Agent Designer is and how we might design a user interface that best enables users to obtain quality results from the system. We show that the Agent Designer is effective for certain musical styles, such as loop-based electronic music, and that we as expert users can design agents that produce the most effective results. We also note that it remains a challenge to automate this process fully.

Backgammon: process-based musical explorations using the agent designer

The Agent Designer is software for designing virtual musical software agents that are capable of remotely controlling a musical composition, making decisions that are derived from previous human performances and additional creative input in the structuring of this example data. This composition, Backgammon, is a collaboration between the Agent Designers creator Aengus Martin and musician Ollie Bown. In this performance, musical agents created with the Agent Designer control part of a dynamic composition in such a way that an improvising partner (Ollie Bown on live electronics) has partial control over what the system can do. In this way the boundary between the active and passive possibilities of software objects becomes the subject of a compositional process. The title refers to the game-like dynamics that arise, involving chance, choices and the ability to dramatically interrupt the others course of action.